Installation¶

jaxfolio targets Python 3.11+ and installs a small, well-scoped dependency set. Optional extras add data loaders and the local-LLM client only when you need them, so the core install stays lean.

Core install¶

uvpip

uv add jaxfolio

pip install jaxfolio

The core install pulls in jax, jaxlib, optax, numpy, pandas, scipy, and matplotlib — everything needed for the optimizers, the backtester, the options toolkit, and the visualizations.

Optional extras¶

jaxfolio splits network- and integration-heavy dependencies into extras. Install only what you use.

Extra	Installs	Enables
`data`	`yfinance`, `pyarrow`	`load_yfinance`, `load_parquet`, `load_option_chain`
`llm`	`requests`	the `OllamaClient` local-model client
`dev`	`pytest`, `pytest-cov`, `ruff`	the test suite and linters

uvpip

uv add "jaxfolio[data]"          # + Yahoo Finance / Parquet loaders
uv add "jaxfolio[llm]"           # + local-model client
uv add "jaxfolio[data,llm]"      # both

pip install "jaxfolio[data]"
pip install "jaxfolio[llm]"
pip install "jaxfolio[data,llm]"

Extras are guarded

Importing jaxfolio never requires an extra. The loaders and the LLM client raise a clear, actionable ImportError only if you call them without the corresponding extra installed — so the core package always imports cleanly.

Verify the installation¶

import jaxfolio as jf

print(jf.__version__)

returns = jf.generate_returns(n_assets=6, seed=0)   # offline synthetic data
result = jf.maximum_sharpe(returns)
print(result)                                       # PortfolioResult(method='Maximum Sharpe', ...)

If that prints a PortfolioResult, you are ready to go — no network access is required, because generate_returns produces reproducible synthetic data locally.

Local LLM prerequisites (optional)¶

The LLM strategies drive a local model via Ollama — no API keys, and no data leaves your machine. After installing the llm extra:

# 1. install Ollama:  https://ollama.com
ollama serve             # start the local server
ollama pull llama3.1     # or mistral, qwen2.5, gemma3, ...

Every LLM strategy also accepts an injected client, so you can run the entire flow offline with the built-in FakeLLM — this is how the tests and examples/04_llm_strategies.py work.

Development setup¶

Clone the repository and sync all extras:

git clone https://github.com/bravant-oss/jaxfolio
cd jaxfolio

uv sync --all-extras
uv run pytest
uv run ruff check . && uv run ruff format --check .

To build these docs locally:

uv sync --extra docs
uv run mkdocs serve          # live preview at http://127.0.0.1:8000
uv run mkdocs build          # render the static site into ./site